Music and podcast work is mostly four file jobs: cut the good take, join the pieces, fix the level, then export the format the platform wants. Each one is a different tool, and only the vocal split uploads anything.
The order is trim, join, normalise, convert. Trim first so you do not carry dead air through everything else. Join second, because merging resamples every file to one common format. Normalise third, so the target loudness is measured on the finished edit. Convert last, once, so a lossy file is not encoded twice for no reason.
1. Cut the take
Audio Trimmer draws the waveform and lets you drag the edges of the region, or type exact start and end times. Because it re-encodes the selection instead of copying compressed frames, the cut lands where you asked rather than on the nearest frame boundary. A fade slider of up to 5 seconds softens both ends, which hides a hard cut on a room recording.
It opens MP3, WAV, M4A, AAC, OGG, FLAC and Opus, and exports MP3 or WAV. MP3 at the default 192 kbps is fine for listening; WAV is the right choice if the file still has to be edited or merged, because it avoids stacking lossy encodes.
The limit is that this is one region per file. Cutting several sections out of a long interview means running it once per section and joining the results afterwards.
Audio Trimmer
Cut a take on the waveform, add a fade of up to 5 seconds, export MP3 or WAV.
2. Join the pieces, in order
Audio Merger takes two or more tracks, orders them with the up and down buttons, and writes one file. Before joining, it converts everything to a common sample rate and to stereo, which is what keeps the seams from jumping between takes recorded on different devices.
Be clear about what this is: a sequence, not a mix. The tracks play one after another. There is no per-track volume, no pan, no crossfade and no way to have a music bed running underneath a voice. If you need that, mix it in a proper editor first and use this tool to join the finished segments.
Merging re-encodes, so a lossy source is encoded once more. Export WAV from this step if the file still has to be normalised or converted.
Merge Audio
Join tracks in the order you set, with one shared sample rate and stereo output.
3. Fix the level
Volume Booster has two modes. The gain slider moves from -20 dB to +20 dB in 0.5 dB steps, and every +6 dB roughly doubles amplitude. Normalize loudness instead measures the average perceived level of the whole track and corrects it to a target following EBU R128 and ITU-R BS.1770, which is what makes two episodes recorded at different levels match. The targets are -14, -16 and -23 LUFS; -16 is a common podcast choice, -14 suits streaming platforms, -23 is the broadcast standard.
Turn on anti-clipping whenever you boost by more than +6 dB or the source already sounds loud. It adds a limiter stage that holds peaks at 0.97 of full scale instead of letting them clip. Output can be MP3, WAV or M4A.
One honest limit: raising gain multiplies the waveform that is already there, so hiss, room noise and muffled speech come up with the voice. Normalisation does not clean a recording, it only matches its level.
Volume Booster
Boost up to +20 dB or normalise to -14, -16 or -23 LUFS with anti-clipping.
4. Export the format
Audio Converter takes up to 30 tracks at a time and converts them one after another, with a download button per file and a Download all option. Input can be MP3, WAV, M4A, AAC, OGG, FLAC or Opus; output can be MP3, WAV, M4A, OGG or FLAC.
Format choice is a compatibility decision, not a quality one. At the same bitrate M4A usually sounds better and produces smaller files, while MP3 plays on more devices and older car stereos. Keep the original sample rate unless the audio is headed for video, where 48000 Hz avoids a resample later. And converting MP3 to WAV never restores what the lossy step discarded; it only gives you an uncompressed copy of what is left.
5. Split vocals, when you need a backing track
Vocal Remover runs a source separation model in the cloud and returns two MP3 files: the vocals and the instrumental. It accepts MP3, WAV, M4A, AAC and FLAC audio, and video in MP4, MOV, MKV or WebM, in which case the audio track is used. One file at a time, up to 30 MB.
This is the only step on the page that leaves your device. The track is uploaded to a server and forwarded to Volcengine AI MediaKit, and the run is metered rather than unlimited. Quality depends on the mix: a clean studio recording separates well, while a dense live recording with reverb bleeds between the two files. Treat the stems as a draft for practice or karaoke, not as a master for release.
Vocal Remover
Split a track into vocals and instrumental. Uploads the file and uses Cloud AI credits.
What this chain cannot do
- No multitrack mixing. Tracks are joined end to end, not layered. There is no per-track volume, pan, crossfade or music bed under a voice.
- No noise reduction, EQ or compression. Normalisation matches loudness; it does not clean up hiss, hum or room echo.
- No real-time loudness meter. You export, check the result, and run it again if the target was missed.
- Merging and trimming re-encode. Plan the chain so the lossy encode happens once, at the end.
- The vocal split is cloud and metered. The file is uploaded, one at a time, up to 30 MB.
- No project files. Nothing here exports stems, automation or an editable session. It exports finished audio.